%matplotlib inline

import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
import plotly.express as px
from scipy import stats
import plotly.graph_objects as go

from glob import glob
from pathlib import Path

files = glob("./open-data-master/data/events/*")

print("Total number of files/events:", len(files))
files[:10]

Total number of files/events: 3433

['./open-data-master/data/events/3879600.json',
 './open-data-master/data/events/3901250.json',
 './open-data-master/data/events/3754300.json',
 './open-data-master/data/events/2275050.json',
 './open-data-master/data/events/19795.json',
 './open-data-master/data/events/3825894.json',
 './open-data-master/data/events/7298.json',
 './open-data-master/data/events/3753996.json',
 './open-data-master/data/events/3900504.json',
 './open-data-master/data/events/3753979.json']

# check for other files than .json
[x for x in files if not x.endswith('.json')]

[]

print("Read file:", files[40])
df_event_test = pd.read_json(files[40]) 

#print("Original dataframe size:", df_event_test.shape)
#print(df_event_test.columns)

#df_event_test = df_event_test[['id', 'play_pattern', 'team', 'pass', 'shot']]
#df_event_test = df_event_test.drop(['timestamp', 'second', 'dribble', 'foul_committed', 'miscontrol', 'foul_won', 'off_camera'], axis=1)

print(f"Size of dataframe: {df_event_test.shape}") # and datatypes:\n{df_event_test.dtypes}") 
df_event_test.head()

Read file: ./open-data-master/data/events/3878553.json
Size of dataframe: (3783, 37)

def get_corner_shot(df_):
    return df_[df_['shot'].apply( # get corners as direct shot if it yields to a goal
            lambda x: (x.get('type', {}).get('id') == 61 if isinstance(x, dict) and 'type' in x else False)
    )].assign(eventcode="direct_corner_shot")

get_corner_shot(df_event_test)

def get_goal_by_corner_shot(df_):
    return df_[df_['shot'].apply( # get corners as direct shot if it yields to a goal
            lambda x: (x.get('type', {}).get('id') == 61 if isinstance(x, dict) and 'type' in x else False) and
                      (x.get('outcome', {}).get('id') == 97 if isinstance(x, dict) and 'outcome' in x else False)
    )].assign(eventcode="direct_goal_via_corner")

get_goal_by_corner_shot(df_event_test)

def get_goal_by_shot_with_previous_corner(df_):
    return df_[(df_['play_pattern'].apply(lambda x: (x.get('id', {}) == 2))) & # check for corner play pattern
               (df_['shot'].apply(lambda x: (x.get('outcome', {}).get('id') == 97 if isinstance(x, dict) and 'outcome' in x else False)))
    ].assign(eventcode="goal_via_shot_with_previous_corner")

get_goal_by_shot_with_previous_corner(df_event_test)

def get_nogoal_by_shot_with_previous_corner(df_):
    return df_[(df_['play_pattern'].apply(lambda x: (x.get('id', {}) == 2))) & # check for corner play pattern
               (df_['shot'].apply(lambda x: (x.get('outcome', {}).get('id') != 97 if isinstance(x, dict) and 'outcome' in x else False)))
    ].assign(eventcode="nogoal_via_shot_with_previous_corner")

get_nogoal_by_shot_with_previous_corner(df_event_test)

def get_corner_passes(df_):
    return df_[df_['pass'].apply(
        lambda x: (x.get('type', {}).get('id') == 61 if isinstance(x, dict) and 'type' in x else False)
    )].assign(eventcode="corner_pass")

get_corner_passes(df_event_test)

def get_corner_pass_for_later_goal(df_):
    
    # get the ids's all events with a goal after a corner 
    idx_goal_event = get_goal_by_shot_with_previous_corner(df_)['index'].values

    # get all events with corner passes
    df_corner_passes = get_corner_passes(df_)

    df_latest_corner_passes = pd.DataFrame()

    # for each index the the latest pass after a corner which is in the underlying playing play pattern of a corner
    for idx in idx_goal_event:
        # get latest corner pass with the playing pattern of the underlying goal
        df_latest_corner_pass = df_corner_passes[df_corner_passes['index'] < idx].sort_values(by='index').tail(1)

        df_latest_corner_passes = pd.concat([df_latest_corner_passes, df_latest_corner_pass])

    return df_latest_corner_passes.assign(eventcode="corner_pass_for_later_goal")

get_corner_pass_for_later_goal(df_event_test)

def collect_information(files):

    df_events = pd.DataFrame()
    
    # load all dataframes by iterating over the files
    for i, file in enumerate(files):  

        # get the filename which corresponds to the match-id
        filename = Path(file).stem

        ### load the current file
        df_event_tmp = pd.read_json(file)  

        # get direct corner shots, add match-id
        df_corner_shots = get_corner_shot(df_event_tmp).assign(match_id=filename)
        
        # get goal via direct corner shots, add match-id
        df_direct_corner_goal = get_goal_by_corner_shot(df_event_tmp).assign(match_id=filename)

        # get goal by shot with previous corner, add match-id
        df_goal_by_shot_with_previous_corner = get_goal_by_shot_with_previous_corner(df_event_tmp).assign(match_id=filename)

        # get corner pass for later goal, add match-id
        df_corner_pass_for_later_goal = get_corner_pass_for_later_goal(df_event_tmp).assign(match_id=filename)
        
        # collect all events together
        df_events = pd.concat([df_events, df_corner_shots, df_direct_corner_goal, df_goal_by_shot_with_previous_corner, df_corner_pass_for_later_goal])
        
        if i % 300 == 0:
            print(f"[{i} / {len(files)}]-Index with match-id: {filename} with Original dataframe size: {df_event_tmp.shape} with collected dataframe size: {df_events.shape}")
            print(f"\tDirect corner shots: {df_corner_shots.shape} ; direct corner goal: {df_direct_corner_goal.shape} ; " 
                + f"goal by shot with previous corner: {df_goal_by_shot_with_previous_corner.shape} ; get corner pass for later goal: {df_corner_pass_for_later_goal.shape}")
    
        #if i == 300:
        #    break
    
    return df_events

%%time

#df_corners = collect_information(files)

#df_corners

def collect_corners_without_goal_play_pattern(files):

    df_corners_wo_goal = pd.DataFrame()
    
    # load all dataframes by iterating over the files
    for i, file in enumerate(files):  

        # get the filename which corresponds to the match-id
        filename = Path(file).stem

        ### load the current file
        df_event_tmp = pd.read_json(file)  

        # get all corner passes
        df_corner_passes = get_corner_passes(df_event_tmp).drop(columns=["eventcode"])
    
        # get the index from all cornes which yiels in a goal
        idx_corner_pass_for_later_goal = get_corner_pass_for_later_goal(df_event_tmp).index

        # remove the corners which results to a goal
        df_corners_without_goal_play_pattern = df_corner_passes.drop(index=idx_corner_pass_for_later_goal)
        
        # collect all events together
        df_corners_wo_goal = pd.concat([df_corners_wo_goal, df_corners_without_goal_play_pattern])
        
        if i % 300 == 0:
            print(f"[{i} / {len(files)}]-Index with match-id: {filename} with Original dataframe size: {df_event_tmp.shape} with collected dataframe size: {df_corners_wo_goal.shape}")
            print(f"\tCorner passes: {df_corner_passes.shape} ; Number of goals via corner per game: {len(idx_corner_pass_for_later_goal)} ; " 
                + f"Resulting corner which does not yield to a goal: {df_corners_without_goal_play_pattern.shape}")
    
        #if i == 300:
        #    break
    
    return df_corners_wo_goal

%%time

#df_corners_wo_goal = collect_corners_without_goal_play_pattern(files)

#df_corners_wo_goal

def collect_shots_with_without_goal_play_pattern_corner(files):

    df_shot_goal = pd.DataFrame()
    
    # load all dataframes by iterating over the files
    for i, file in enumerate(files):  

        # get the filename which corresponds to the match-id
        filename = Path(file).stem

        ### load the current file
        df_shot_tmp = pd.read_json(file)  

        # get shots which yields a goal with play pattern "From corner"
        df_shot_goal_tmp = get_goal_by_shot_with_previous_corner(df_shot_tmp)

        # get shots which yields no goal with play pattern "From corner"
        df_shot_nogoal_tmp = get_nogoal_by_shot_with_previous_corner(df_shot_tmp)
        
        # collect all events together
        df_shot_goal = pd.concat([df_shot_goal, df_shot_goal_tmp, df_shot_nogoal_tmp])
        
        if i % 300 == 0:
            print(f"[{i} / {len(files)}]-Index with match-id: {filename} with Original dataframe size: {df_shot_tmp.shape} with collected dataframe size: {df_shot_goal.shape}")
            print(f"\tShot to goal: {df_shot_goal_tmp.shape} ; Shots to no goal: {df_shot_nogoal_tmp.shape}")
    
        #if i == 300:
        #    break
    
    return df_shot_goal

%%time

#df_shot_goal_nogoals = collect_shots_with_without_goal_play_pattern_corner(files)

#df_shot_goal_nogoals

df_shot_goal_nogoals_edited = (df_shot_goal_nogoals    
    .assign(goal=lambda df_: df_['eventcode'].apply(lambda x: "Yes" if x == "goal_via_shot_with_previous_corner" else "No"))   
)[["id", "goal"]] 

df_shot_goal_nogoals_edited

#df_corners.to_json("./intermediate_datasets/corners_intermediate.json", orient='records')

#df_corners_wo_goal.to_json("./intermediate_datasets/corners_wo_goal_intermediate.json", orient='records')

#df_shot_goal_nogoals_edited.to_csv("./intermediate_datasets/shots_after_corner_w_wo_goal_intermediate.csv")

df_corners = pd.read_json("./intermediate_datasets/corners_intermediate.json")

print(f"DataFrame size: {df_corners.shape}")
df_corners.head()

DataFrame size: (2458, 44)

df_corners_wo_goal = pd.read_json("./intermediate_datasets/corners_wo_goal_intermediate.json")

print(f"DataFrame size: {df_corners_wo_goal.shape}")
df_corners_wo_goal.head()

DataFrame size: (32536, 42)

fig = px.histogram(df_corners, x='eventcode')
fig.show()

# get corner pass technique
df_corner_pass_for_later_goal = (df_corners
    .query("eventcode == 'corner_pass_for_later_goal'")
    .assign(corner_pass_technique=lambda df_: df_['pass'].apply(lambda x: x.get('technique', {}).get('name') if isinstance(x, dict) and 'technique' in x else np.nan))
    .assign(corner_pass_height=lambda df_: df_['pass'].apply(lambda x: x.get('height', {}).get('name') if isinstance(x, dict) and 'height' in x else np.nan))
    .assign(pass_length=lambda df_: df_['pass'].apply(lambda x: x.get('length', {}) if isinstance(x, dict) and 'length' in x else np.nan))
    .assign(pass_angle=lambda df_: df_['pass'].apply(lambda x: x.get('angle', {}) if isinstance(x, dict) and 'angle' in x else np.nan))  
    .assign(end_loc_x=lambda df_: df_['pass'].apply(lambda x: x.get('end_location', {})[0] if isinstance(x, dict) and 'end_location' in x else np.nan)) 
    .assign(end_loc_y=lambda df_: df_['pass'].apply(lambda x: x.get('end_location', {})[1] if isinstance(x, dict) and 'end_location' in x else np.nan)) 
    .dropna(subset=['corner_pass_technique', 'corner_pass_height', 'pass_length', 'pass_angle'])
    .rename(columns={
        "corner_pass_technique": "Corner-pass technique",
        "corner_pass_height": "Corner-pass height",
        "pass_length": "Pass length",
        "pass_angle": "Pass angle",
    })
    .assign(goal="Yes")
) 

print(f"Number of corner passes: {df_corner_pass_for_later_goal.shape}")
df_corner_pass_for_later_goal.head()

Number of corner passes: (946, 51)

df_corners_wo_goal_intermediate = (df_corners_wo_goal
    .assign(corner_pass_technique=lambda df_: df_['pass'].apply(lambda x: x.get('technique', {}).get('name') if isinstance(x, dict) and 'technique' in x else np.nan))
    .assign(corner_pass_height=lambda df_: df_['pass'].apply(lambda x: x.get('height', {}).get('name') if isinstance(x, dict) and 'height' in x else np.nan))
    .assign(pass_length=lambda df_: df_['pass'].apply(lambda x: x.get('length', {}) if isinstance(x, dict) and 'length' in x else np.nan))
    .assign(pass_angle=lambda df_: df_['pass'].apply(lambda x: x.get('angle', {}) if isinstance(x, dict) and 'angle' in x else np.nan))
    .dropna(subset=['corner_pass_technique', 'corner_pass_height', 'pass_length', 'pass_angle'])
    .rename(columns={
        "corner_pass_technique": "Corner-pass technique",
        "corner_pass_height": "Corner-pass height",
        "pass_length": "Pass length",
        "pass_angle": "Pass angle",
    })
    .assign(goal="No")                             
)

df_corners_wo_goal_intermediate

df_corner_direct_goal = (df_corners
    .groupby(["eventcode"]).count()[["id"]]
    .reset_index()
    .rename(columns={"eventcode": "Event", "id": "Count"})
    .replace(['direct_corner_shot', 'direct_goal_via_corner'], ['Direct corner shot', 'Direct corner shot with goal'])
    .query(f"Event == 'Direct corner shot' or Event == 'Direct corner shot with goal'")
)

df_corner_direct_goal

# plot only the direct corners
fig = px.histogram(df_corner_direct_goal, x="Event", y="Count", color="Event", height=500, color_discrete_sequence=px.colors.qualitative.Plotly)

# x-axis
fig.update_xaxes(
    title="", #"Shot that leads to a goal",
    title_font_size=30,
    tickfont=dict(size=20),   
    title_standoff=20,
)

# y-axis
fig.update_yaxes(
    title="Count",
    title_font_size=30,
    tickfont=dict(size=20),
    title_standoff=10
)

# legend
fig.update_layout(
    legend=dict(
    yanchor="top",
    y=0.99,
    font=dict(size=20))
)

fig.update_layout(
    showlegend=False,
    font={'size': 25},
    title={'text': '<b>Direct corners with/without goal</b>', 'font': {'size': 30}}
)

fig.show()

fig.write_html("./figures/Corner_shot_w-wo_goal.html", include_plotlyjs="cdn")

# plot the corner pass technique with according pass height
fig = px.histogram(df_corner_pass_for_later_goal, x="Corner-pass technique", color="Corner-pass height",
            height=700, width=1200, color_discrete_sequence=px.colors.qualitative.D3, barmode="group")

# y-axis
fig.update_yaxes(
    title="Count",
    title_font_size=30,
    tickfont=dict(size=20),
    title_standoff=10
)

# legend
fig.update_layout(
    legend=dict(
    yanchor="top",
    y=0.99,
    font=dict(size=30))
)

fig.update_layout(
    font={'size': 25},
    title={'text': '<b>Corner pass technique and pass height</b>', 'font': {'size': 30}}
)

fig.show()

fig.write_html("./figures/Corner_pass_technique_height.html", include_plotlyjs="cdn")

df_goal = df_corner_pass_for_later_goal[['id', 'Corner-pass technique', 'Corner-pass height', 'Pass length', 'Pass angle', 'goal']]
df_goal.head()

df_nogoal = df_corners_wo_goal_intermediate[['id', 'Corner-pass technique', 'Corner-pass height', 'Pass length', 'Pass angle', 'goal']]
df_nogoal.head()

df_w_wo_goal = pd.concat([df_goal, df_nogoal])

#df_w_wo_goal.to_json("./intermediate_datasets/corners_w_wo_goal_statistics.json", orient='records')

df_w_wo_goal = pd.read_json("./intermediate_datasets/corners_w_wo_goal_statistics.json")

print(f"DataFrame size with goals: {df_goal.shape} without goals: {df_nogoal.shape} and in total: {df_w_wo_goal.shape}")
df_w_wo_goal.head()

DataFrame size with goals: (946, 6) without goals: (24945, 6) and in total: (25891, 6)

fig = px.box(df_w_wo_goal, y=["Pass length", "Pass angle"], color="goal", facet_col="Corner-pass technique",
             color_discrete_sequence=px.colors.qualitative.Set1, points="outliers", width=1200, height=700)

# x-axis
fig.update_xaxes(
    title="",
    tickfont=dict(size=20),
    zerolinewidth=2, 
    zerolinecolor='black'   
)

# y-axis
fig.update_yaxes(
    title="Value",
    title_font_size=30,
    tickfont=dict(size=20),
    dtick=5    
)

# legend
fig.update_layout(
    boxgap=0.4,
    boxgroupgap=0.4,
    legend=dict(
        yanchor="top",
        font=dict(size=25),
    )
)

# lines width and marker size
fig.update_traces(
    line=dict(width=3.),
    marker={'size': 7}
)

# add mean
fig.for_each_trace(lambda trace: trace.update(boxmean=True))     

# facet label
fig.for_each_annotation(lambda a: a.update(font=dict(size=20), text=a.text.split("=")[-1]))

# remove second and third yaxis title
fig.update_layout(yaxis2=dict(title=""), yaxis3=dict(title=""))

fig.show()

fig.write_html("./figures/Corner_pass_w_wo_goal_box.html", include_plotlyjs="cdn")

# test for significant differences with regard to the pass length for corner pass 'Inswinging' technique wheater this corner technique results in a goal or not
arr_no_goal = df_nogoal[df_nogoal['Corner-pass technique']=='Inswinging']['Pass length'].to_numpy()

print(len(arr_no_goal), np.var(arr_no_goal))
arr_no_goal

12239 67.14898590683532

array([36.289116, 40.9518  , 32.773922, ..., 36.656513, 39.171036,
       40.00925 ])

arr_goal = df_goal[df_goal['Corner-pass technique']=='Inswinging']['Pass length'].to_numpy()

print(len(arr_goal), np.var(arr_goal))
arr_goal[:10]

470 35.72559864806411

array([32.33945 , 36.122154, 50.243904, 47.420353, 32.125534, 33.306156,
       39.870415, 35.20568 , 37.981705, 43.304157])

# Get the p-value
statistic, pvalue = stats.ttest_ind(
    arr_no_goal,
    arr_goal,
    equal_var=False,
)#[1]

print(statistic, pvalue)

0.07434242135502345 0.9407655102236525

fig = px.box(df_w_wo_goal, y=["Pass length", "Pass angle"], color="Corner-pass technique", 
             color_discrete_sequence=px.colors.qualitative.Dark2, points="outliers", width=1200, height=700)

# x-axis
fig.update_xaxes(
    title="",
    tickfont=dict(size=20),
    zerolinewidth=2, 
    zerolinecolor='black'   
)

# y-axis
fig.update_yaxes(
    title="Value",
    title_font_size=30,
    tickfont=dict(size=20),
    dtick=5    
)

# legend
fig.update_layout(
    boxgap=0.4,
    boxgroupgap=0.4,
    legend=dict(
        yanchor="top",
        font=dict(size=25),
    )
)

# lines width and marker size
fig.update_traces(
    line=dict(width=3.),
    marker={'size': 7}
)

# add mean
fig.for_each_trace(lambda trace: trace.update(boxmean=True))        

fig.show()

fig.write_html("./figures/Corner_pass_length_and_angle_box.html", include_plotlyjs="cdn")

fig = px.scatter_polar(df_corner_pass_for_later_goal, theta="Pass angle", color="Corner-pass technique", range_theta=[30,150])

fig.update_layout(
    polar = dict(radialaxis = dict(showticklabels=False, ticks=''))
)

# legend
fig.update_layout(
    legend=dict(
    yanchor="top",
    y=0.99,
    x=0.1,
    font=dict(size=20))
)

fig.show()

fig.write_html("./figures/Corner_pass_angle_polar.html", include_plotlyjs="cdn")

# determine the rectangle for the individual corner pass technique
x0_inswing = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Inswinging']['end_loc_x'].min()
x1_inswing = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Inswinging']['end_loc_x'].max()
y0_inswing = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Inswinging']['end_loc_y'].min()
y1_inswing = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Inswinging']['end_loc_y'].max()

x0_outswing = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Outswinging']['end_loc_x'].min()
x1_outswing = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Outswinging']['end_loc_x'].max()
y0_outswing = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Outswinging']['end_loc_y'].min()
y1_outswing = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Outswinging']['end_loc_y'].max()

x0_straight = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Straight']['end_loc_x'].min()
x1_straight = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Straight']['end_loc_x'].max()
y0_straight = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Straight']['end_loc_y'].min()
y1_straight = df_corner_pass_for_later_goal[df_corner_pass_for_later_goal['Corner-pass technique']=='Straight']['end_loc_y'].max()

fig = px.scatter(df_corner_pass_for_later_goal, x="end_loc_x", y="end_loc_y", width=1300, height=800, 
                 color="Corner-pass technique", color_discrete_sequence=px.colors.qualitative.Safe, trendline="ols") # ols, lowess

# x-axis
fig.update_xaxes(
    title="Soccer pitch x-axis",
    title_font_size=25,
    tickfont=dict(size=20),
    dtick=10,
    range=[0, 120.8],
    zerolinewidth=2, 
    zerolinecolor='black',
)

# y-axis
fig.update_yaxes(
    title="Soccer pitch y-axis",
    title_font_size=25,
    tickfont=dict(size=20),
    range=[0, 80],
    zerolinewidth=2, 
    zerolinecolor='black',
)

# lines width and marker size
fig.update_traces(marker={'size': 7})

# legend
fig.update_layout(
    legend=dict(
    yanchor="top",
    y=0.99,
    xanchor="left",
    font=dict(size=25),
))

# Update the trendline thickness (modify the first trace, which is the trendline)
fig.update_traces(line=dict(width=5), selector=dict(type='scatter', mode='lines'))

fig.update_layout(
    font={'size': 25},
    title={'text': '<b>End location of corner pass before goal play pattern</b>', 'font': {'size': 30}}
)

# Add rectangle for Inswinging corner
fig.add_shape(type="rect",
    xref="x", yref="y",
    x0=x0_inswing, y0=y0_inswing,
    x1=x1_inswing, y1=y1_inswing,
    opacity=0.2,
    fillcolor=px.colors.qualitative.Safe[0],
    line_color=px.colors.qualitative.Safe[0],
)

# Add rectangle for Outswinging corner
fig.add_shape(type="rect",
    xref="x", yref="y",
    x0=x0_outswing, y0=y0_outswing,
    x1=x1_outswing, y1=y1_outswing,
    opacity=0.2,
    fillcolor=px.colors.qualitative.Safe[1],
    line_color=px.colors.qualitative.Safe[1],
)

# Add rectangle for Straigt corner
fig.add_shape(type="rect",
    xref="x", yref="y",
    x0=x0_straight, y0=y0_straight,
    x1=x1_straight, y1=y1_straight,
    opacity=0.15,
    fillcolor=px.colors.qualitative.Safe[2],
    line_color=px.colors.qualitative.Safe[2],
)

fig.show()

fig.write_html("./figures/Corner_pass_end_location.html", include_plotlyjs="cdn")

files_freese_frames = glob("./open-data-master/data/three-sixty/*")

print("Total number of files/360 sets:", len(files_freese_frames))
files[:10]

Total number of files/360 sets: 295

['./open-data-master/data/three-sixty/3902240.json',
 './open-data-master/data/three-sixty/3938644.json',
 './open-data-master/data/three-sixty/3788742.json',
 './open-data-master/data/three-sixty/3893798.json',
 './open-data-master/data/three-sixty/3869354.json',
 './open-data-master/data/three-sixty/3893809.json',
 './open-data-master/data/three-sixty/3788754.json',
 './open-data-master/data/three-sixty/3893825.json',
 './open-data-master/data/three-sixty/3895158.json',
 './open-data-master/data/three-sixty/3857292.json']

print("Read file:", files_freese_frames[40])
df_three_test = pd.read_json(files_freese_frames[40]) 

print(df_three_test.columns)

print(f"Size of dataframe: {df_three_test.shape}") # and datatypes:\n{df_event_test.dtypes}") 
df_three_test.head()

Read file: ./open-data-master/data/three-sixty/3857266.json
Index(['event_uuid', 'visible_area', 'freeze_frame'], dtype='object')
Size of dataframe: (3290, 3)

df_three_test_single = df_three_test.loc[42:43]
df_three_test_single

def get_visible_area_coordinates(df_):  
    return (pd.DataFrame()
        .assign(visible_area_x=df_['visible_area'].apply(lambda x: x[::2]))
        .assign(visible_area_y=df_['visible_area'].apply(lambda x: x[1::2]))
        .assign(event_uuid=df_['event_uuid'].apply(lambda x: x))
        .set_index('event_uuid').apply(pd.Series.explode).reset_index()        
    )

get_visible_area_coordinates(df_three_test_single)

def get_freeze_frame_info(df_):
    return (pd.DataFrame()
        .assign(teammate=df_['freeze_frame'].apply(lambda x: [player['teammate'] for player in x]))
        .assign(actor=df_['freeze_frame'].apply(lambda x: [player['actor'] for player in x]))         
        .assign(keeper=df_['freeze_frame'].apply(lambda x: [player['keeper'] for player in x]))
        .assign(location_x=df_['freeze_frame'].apply(lambda x: [player['location'][0] for player in x])) 
        .assign(location_y=df_['freeze_frame'].apply(lambda x: [player['location'][1] for player in x])) 
        .assign(event_uuid=df_['event_uuid'].apply(lambda x: x))
        .set_index('event_uuid').apply(pd.Series.explode).reset_index()
        .assign(teammate_category=lambda x: ['teammate' if val == True else 'opponent' for val in x['teammate']])
        .assign(actor_category=lambda x: ['actor' if val == True else 'non-actor' for val in x['actor']])
        .assign(keeper_category=lambda x: ['keeper' if val == True else 'player' for val in x['keeper']])
    )

get_freeze_frame_info(df_three_test_single)

df_three_test_single_visible_area = get_visible_area_coordinates(df_three_test_single)
df_three_test_single_visible_area = df_three_test_single_visible_area[df_three_test_single_visible_area["event_uuid"]=="2439c812-71f9-4fc1-864b-61fe7892add7"]
df_three_test_single_visible_area

df_three_test_single_freeze_frame = get_freeze_frame_info(df_three_test_single)
df_three_test_single_freeze_frame = df_three_test_single_freeze_frame[df_three_test_single_freeze_frame["event_uuid"]=="2439c812-71f9-4fc1-864b-61fe7892add7"]
df_three_test_single_freeze_frame

polygon_x = df_three_test_single_visible_area['visible_area_x'].to_numpy()
polygon_y = df_three_test_single_visible_area['visible_area_y'].to_numpy()

print(polygon_x, polygon_y)

[0.0 0.0 30.3246902472383 61.3426654740077 61.0777090348261 0.0] [80.0 61.6619136535695 9.00661251365509 18.3903590167589 80.0 80.0]

fig = px.scatter(df_three_test_single_freeze_frame, x="location_x", y="location_y", color="teammate_category", symbol="actor_category", 
                 symbol_sequence=['star', 'circle'], color_discrete_sequence=px.colors.qualitative.Safe, width=1200, height=600)#, trendline="ols") # ols, lowess

# x-axis
fig.update_xaxes(
    title="Soccer pitch x-axis",
    title_font_size=25,
    tickfont=dict(size=20),
    dtick=10,
    range=[0, 120.8],
    zerolinewidth=2, 
    zerolinecolor='black',
)

# y-axis
fig.update_yaxes(
    title="Soccer pitch y-axis",
    title_font_size=25,
    tickfont=dict(size=20),
    range=[0, 80],
    zerolinewidth=2, 
    zerolinecolor='black',
)

# Add the polygon for the visible area
fig.add_trace(go.Scatter(
    x=polygon_x, 
    y=polygon_y, 
    #fill='toself',                         
    mode='lines', 
    line=dict(color='orange', width=5), 
    fillcolor='lightblue', 
    name='Visible area'),
)

# lines width and marker size
fig.update_traces(marker={'size': 12})

# legend
fig.update_layout(   
    legend=dict(
         title="Player category",
        yanchor="top",
        y=0.99,
        xanchor="left",
        font=dict(size=25),
))

# Update the polygon selector of the visible area (modify the first trace, which is the polygon)
fig.update_traces(selector=dict(type='scatter', mode='lines'))

fig.update_layout(
    font={'size': 25},
    title={'text': '<b>Freeze frame with outlined visible area and player position</b>', 'font': {'size': 30}}
)

fig.show()

fig.write_html("./figures/Freeze_frame_game_sitation_single.html", include_plotlyjs="cdn")

def collect_visible_areas_and_freeze_frames(files_freese_frames):

    df_visible_areas = pd.DataFrame()
    df_freeze_frames = pd.DataFrame()
    
    # load all dataframes by iterating over the files
    for i, file in enumerate(files_freese_frames):  

        # get the filename which corresponds to the match-id (not mentioned in the documentation)
        filename = Path(file).stem

        ### load the current file
        try:
            df_tmp = pd.read_json(file)  
        except:
            print(f"ERROR [{i} / {len(files)}]-Index with match-id: {filename}")
        
        # get all visible areas
        df_visible_area_tmp = get_visible_area_coordinates(df_tmp)

        # get all freeze frames
        df_freeze_frame_tmp = get_freeze_frame_info(df_tmp)
        
        # collect all visible areas together
        df_visible_areas = pd.concat([df_visible_areas, df_visible_area_tmp])

        # collect all freeze frames together
        df_freeze_frames = pd.concat([df_freeze_frames, df_freeze_frame_tmp])
        
        if i % 50 == 0:
            print(f"[{i} / {len(files)}]-Index with match-id: {filename} with Original dataframe size: {df_tmp.shape}")
            print(f"\tvisible areas: {df_visible_areas.shape} ; freeze frames: {df_freeze_frames.shape}")
    
        #if i == 5:
        #    break
    
    return df_visible_areas, df_freeze_frames

%%time

df_visible_areas, df_freeze_frames = collect_visible_areas_and_freeze_frames(files_freese_frames)

df_visible_areas

[0 / 295]-Index with match-id: 3902240 with Original dataframe size: (3811, 3)
	visible areas: (20802, 3) ; freeze frames: (53990, 9)
ERROR [14 / 295]-Index with match-id: 3835338
[50 / 295]-Index with match-id: 3895250 with Original dataframe size: (3410, 3)
	visible areas: (907980, 3) ; freeze frames: (2403578, 9)
ERROR [73 / 295]-Index with match-id: 3835342
[100 / 295]-Index with match-id: 3902968 with Original dataframe size: (3373, 3)
	visible areas: (1841497, 3) ; freeze frames: (4857383, 9)
[150 / 295]-Index with match-id: 3788773 with Original dataframe size: (3248, 3)
	visible areas: (2761443, 3) ; freeze frames: (7327130, 9)
ERROR [198 / 295]-Index with match-id: 3845506
[200 / 295]-Index with match-id: 3857254 with Original dataframe size: (3157, 3)
	visible areas: (3692270, 3) ; freeze frames: (9852184, 9)
[250 / 295]-Index with match-id: 3930169 with Original dataframe size: (2815, 3)
	visible areas: (4594949, 3) ; freeze frames: (12208682, 9)
CPU times: user 2min 57s, sys: 1min 15s, total: 4min 13s
Wall time: 4min 39s

df_visible_areas.head()

df_freeze_frames.head()

df_shot_goal_nogoals_edited = pd.read_csv("./intermediate_datasets/shots_after_corner_w_wo_goal_intermediate.csv", index_col=0)

print(f"DataFrame size with and without goals: {df_shot_goal_nogoals_edited.shape}")
df_shot_goal_nogoals_edited.head()

DataFrame size with and without goals: (13924, 2)

df_freeze_frame_event = (df_freeze_frames
    .merge(df_shot_goal_nogoals_edited, how="left", left_on="event_uuid", right_on="id")
    .dropna(subset=["goal"])
    .drop(columns=["id"])
    .assign(marker_size=lambda df_: df_['actor'].apply(lambda x: 0.4 if x == True else 0.07))           
)

df_freeze_frame_event

fig = px.scatter(df_freeze_frame_event, x="location_x", y="location_y", color="teammate_category", symbol="actor_category", 
                 facet_col="goal", facet_col_wrap=4, facet_col_spacing=0.03, symbol_sequence=['circle', 'star'], size="marker_size",
                 color_discrete_sequence=px.colors.qualitative.Safe, width=1600, height=600)#, trendline="ols") # ols, lowess

# x-axis
fig.update_xaxes(
    title="Soccer pitch x-axis",
    title_font_size=25,
    tickfont=dict(size=20),
    dtick=10,
    range=[0, 120.8],
    zerolinewidth=2, 
    zerolinecolor='black',
)

# y-axis
fig.update_yaxes(
    title="Soccer pitch y-axis",
    title_font_size=25,
    tickfont=dict(size=20),
    range=[0, 80],
    zerolinewidth=2, 
    zerolinecolor='black',
)

# lines width and marker size
#fig.update_traces(marker={'size': 5})

# legend
fig.update_layout(   
    legend=dict(
         title="Player category",
        yanchor="top",
        y=0.99,
        xanchor="left",
        font=dict(size=25),
))

# Update the polygon selector of the visible area (modify the first trace, which is the polygon)
fig.update_traces(selector=dict(type='scatter', mode='lines'))

# facet label
fig.for_each_annotation(lambda a: a.update(font=dict(size=20), text=a.text.split("=")[-1]))

# remove second yaxis title
fig.update_layout(yaxis2=dict(title=""))

fig.update_layout(
    font={'size': 25},
    title={'text': '<b>Freeze frame player position resulting in goal/no goal</b>', 'font': {'size': 30}}
)

fig.show()

fig.write_html("./figures/Freeze_frame_game_sitation_goal_vs_nogoal.html", include_plotlyjs="cdn")

from scipy.spatial import ConvexHull

points_teammate_goal = df_freeze_frame_event.query("teammate_category == 'teammate' and goal == 'Yes' and actor_category == 'actor'")[["location_x", "location_y"]].to_numpy()
points_teammate_nogoal = df_freeze_frame_event.query("teammate_category == 'teammate' and goal == 'No' and actor_category == 'actor'")[["location_x", "location_y"]].to_numpy()

points_opponent_goal = df_freeze_frame_event.query("teammate_category == 'opponent' and goal == 'Yes'")[["location_x", "location_y"]].to_numpy()
points_opponent_nogoal = df_freeze_frame_event.query("teammate_category == 'opponent' and goal == 'No'")[["location_x", "location_y"]].to_numpy()

print(points_teammate_goal.shape, points_teammate_nogoal.shape)
print(points_opponent_goal.shape, points_opponent_nogoal.shape)

(109, 2) (1190, 2)
(1096, 2) (12267, 2)

polygon_teammate_goal = points_teammate_goal[ConvexHull(points_teammate_goal).vertices]
polygon_teammate_nogoal = points_teammate_nogoal[ConvexHull(points_teammate_nogoal).vertices]

polygon_opponent_goal = points_opponent_goal[ConvexHull(points_opponent_goal).vertices]
polygon_opponent_nogoal = points_opponent_nogoal[ConvexHull(points_opponent_nogoal).vertices]

print(polygon_teammate_goal.shape, polygon_teammate_nogoal.shape)
print(polygon_opponent_goal.shape, polygon_opponent_nogoal.shape)

(6, 2) (6, 2)
(11, 2) (11, 2)

polygon_teammate_x = points_teammate[hull_teammate.vertices][:, 0]
polygon_teammate_y = points_teammate[hull_teammate.vertices][:, 1]

fig = px.scatter(df_freeze_frame_event, x="location_x", y="location_y", color="teammate_category", symbol="actor_category", 
                 facet_col="goal", facet_col_wrap=4, facet_col_spacing=0.03, symbol_sequence=['circle', 'star'], 
                 size="marker_size", color_discrete_sequence=px.colors.qualitative.Safe, width=1600, height=600, trendline="ols") # ols, lowess

# x-axis
fig.update_xaxes(
    title="Soccer pitch x-axis",
    title_font_size=25,
    tickfont=dict(size=20),
    dtick=10,
    range=[0, 120.8],
    zerolinewidth=2, 
    zerolinecolor='black',
)

# y-axis
fig.update_yaxes(
    title="Soccer pitch y-axis",
    title_font_size=25,
    tickfont=dict(size=20),
    range=[0, 80],
    zerolinewidth=2, 
    zerolinecolor='black',
)

# lines width and marker size
#fig.update_traces(marker={'size': 5})

# legend
fig.update_layout(   
    legend=dict(
         title="Player category",
        yanchor="top",
        y=0.99,
        xanchor="left",
        font=dict(size=25),
))

# Add polygon to the 1st facet (opponent team)
fig.add_trace(
    go.Scatter(
        x=polygon_opponent_nogoal[:, 0], # x
        y=polygon_opponent_nogoal[:, 1], # y
        mode='lines',
        fill='toself',
        fillcolor=px.colors.qualitative.Safe[0],
        line=dict(color=px.colors.qualitative.Safe[0]),
        opacity=0.2,
        name='Opponent',
        showlegend=True,  # Hide from legend
        hoverinfo='skip'
    ),
    row=1, col=1 
)

# Add polygon to the 1st facet (teammates)
fig.add_trace(
    go.Scatter(
        x=polygon_teammate_nogoal[:, 0], # x
        y=polygon_teammate_nogoal[:, 1], # y
        mode='lines',
        fill='toself',
        fillcolor=px.colors.qualitative.Safe[1],
        line=dict(color=px.colors.qualitative.Safe[1]),
        opacity=0.2,
        name='Teammates',
        showlegend=True,  # Hide from legend
        hoverinfo='skip'
    ),
    row=1, col=1 
)

# Add polygon to the 2nd facet (opponent team)
fig.add_trace(
    go.Scatter(
        x=polygon_opponent_goal[:, 0], # x
        y=polygon_opponent_goal[:, 1], # y
        mode='lines',
        fill='toself',
        fillcolor=px.colors.qualitative.Safe[0],
        line=dict(color=px.colors.qualitative.Safe[0]),
        opacity=0.2,
        name='Opponent',
        showlegend=False,  # Hide from legend
        hoverinfo='skip'
    ),
    row=1, col=2 
)

# Add polygon to the 2nd facet (teammates)
fig.add_trace(
    go.Scatter(
        x=polygon_teammate_goal[:, 0], # x
        y=polygon_teammate_goal[:, 1], # y
        mode='lines',
        fill='toself',
        fillcolor=px.colors.qualitative.Safe[1],
        line=dict(color=px.colors.qualitative.Safe[1]),
        opacity=0.2,
        name='Teammates',
        showlegend=False,  # Hide from legend
        hoverinfo='skip'
    ),
    row=1, col=2 
)

# Update the polygon selector of the visible area (modify the first trace, which is the polygon)
fig.update_traces(selector=dict(type='scatter', mode='lines'))

# facet label
fig.for_each_annotation(lambda a: a.update(font=dict(size=20), text=a.text.split("=")[-1]))

# remove second yaxis title
fig.update_layout(yaxis2=dict(title=""))

fig.update_layout(
    font={'size': 25},
    title={'text': '<b>Freeze frame team area and running direction resulting in goal/no goal</b>', 'font': {'size': 30}}
)

fig.show()

fig.write_html("./figures/Freeze_frame_game_sitation_goal_vs_nogoal_polygon_trend.html", include_plotlyjs="cdn")

#df_freeze_frame_event.to_csv("./intermediate_datasets/freeze_frames_event.csv")
##df_freeze_frame_event.to_json("./intermediate_datasets/freeze_frames_df_freeze_frame_event.json", orient='records')

df_freeze_frame_event = pd.read_csv("./intermediate_datasets/freeze_frames_event.csv", index_col=0)

df_freeze_frame_event['y'] = df_freeze_frame_event['goal'].apply(lambda x: 1 if x == 'Yes' else 0)

print(f"DataFrame size: {df_freeze_frame_event.shape}")
df_freeze_frame_event.head()

DataFrame size: (22887, 12)

np.unique(df_freeze_frame_event.goal, return_counts=True)

(array(['No', 'Yes'], dtype=object), array([21024,  1863]))

from sklearn.metrics import accuracy_score, f1_score, precision_score, recall_score, classification_report, confusion_matrix
from sklearn.model_selection import train_test_split
from sklearn.pipeline import make_pipeline

from sklearn.ensemble import RandomForestClassifier
from sklearn.gaussian_process.kernels import RBF
from sklearn.naive_bayes import GaussianNB
from sklearn.neighbors import KNeighborsClassifier
from sklearn.neural_network import MLPClassifier
from sklearn.svm import SVC
from xgboost import XGBClassifier

# split in features and labels
#X = df_freeze_frame_event[['actor', 'teammate', 'keeper', 'location_x', 'location_y']]
X = df_freeze_frame_event[['actor', 'location_x', 'location_y']]
y = df_freeze_frame_event['y']

# split the data in training and test data
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=42)

print(f"X: {X.shape} ; y: {y.shape}")
print(f"X_train: {X_train.shape} ; y_train: {y_train.shape}")
print(f"X_test: {X_test.shape} ; y_test: {y_test.shape}")

X: (22887, 3) ; y: (22887,)
X_train: (16020, 3) ; y_train: (16020,)
X_test: (6867, 3) ; y_test: (6867,)

# define the classifiers to be tested
names = [
    "Nearest Neighbors",
    "RBF SVM",
    "XGBoost",
    "Random Forest",
    "Neural Net",
    "Naive Bayes",
]

classifiers = [
    KNeighborsClassifier(3),
    SVC(random_state=42),
    XGBClassifier(n_estimators=100, max_depth=100, learning_rate=1, objective='binary:logistic'),
    RandomForestClassifier(
        max_depth=100, n_estimators=100, random_state=42
    ),
    MLPClassifier(max_iter=1000, random_state=42),
    GaussianNB()
]

%%time

# iterate over classifiers
for name, clf in zip(names, classifiers):   

    clf = make_pipeline(StandardScaler(), clf)
    clf.fit(X_train, y_train)

    print("Current classifier:", name)
    
    # predict the test data
    y_pred = clf.predict(X_test)
    
    # and evaluate the results evaluate
    print("Accuracy:", accuracy_score(y_test, y_pred))
    print(classification_report(y_test, y_pred))
    #print(confusion_matrix(y_test, y_pred))

Current classifier: Nearest Neighbors
Accuracy: 0.9027231687782147
              precision    recall  f1-score   support

           0       0.92      0.98      0.95      6317
           1       0.12      0.03      0.05       550

    accuracy                           0.90      6867
   macro avg       0.52      0.51      0.50      6867
weighted avg       0.86      0.90      0.88      6867

Current classifier: RBF SVM
Accuracy: 0.9199068006407456
              precision    recall  f1-score   support

           0       0.92      1.00      0.96      6317
           1       0.00      0.00      0.00       550

    accuracy                           0.92      6867
   macro avg       0.46      0.50      0.48      6867
weighted avg       0.85      0.92      0.88      6867

Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.

Current classifier: XGBoost
Accuracy: 0.8814620649483035
              precision    recall  f1-score   support

           0       0.92      0.95      0.94      6317
           1       0.14      0.09      0.11       550

    accuracy                           0.88      6867
   macro avg       0.53      0.52      0.52      6867
weighted avg       0.86      0.88      0.87      6867

Current classifier: Random Forest
Accuracy: 0.9169943206640454
              precision    recall  f1-score   support

           0       0.92      0.99      0.96      6317
           1       0.33      0.04      0.07       550

    accuracy                           0.92      6867
   macro avg       0.63      0.52      0.51      6867
weighted avg       0.87      0.92      0.89      6867

Current classifier: Neural Net
Accuracy: 0.9199068006407456
              precision    recall  f1-score   support

           0       0.92      1.00      0.96      6317
           1       0.00      0.00      0.00       550

    accuracy                           0.92      6867
   macro avg       0.46      0.50      0.48      6867
weighted avg       0.85      0.92      0.88      6867

Current classifier: Naive Bayes
Accuracy: 0.9199068006407456
              precision    recall  f1-score   support

           0       0.92      1.00      0.96      6317
           1       0.00      0.00      0.00       550

    accuracy                           0.92      6867
   macro avg       0.46      0.50      0.48      6867
weighted avg       0.85      0.92      0.88      6867

CPU times: user 13.8 s, sys: 1.73 s, total: 15.5 s
Wall time: 4.6 s

Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.

from IPython.display import display, HTML

with open('./figures/Corner_shot_w-wo_goal.html', 'r') as file:
    Corner_shot_w_wo_goal_html = file.read()

with open('./figures/Corner_pass_technique_height.html', 'r') as file:
    Corner_pass_technique_height = file.read()

with open('./figures/Corner_pass_w_wo_goal_box.html', 'r') as file:
    Corner_pass_w_wo_goal_box = file.read()

with open('./figures/Corner_pass_length_and_angle_box.html', 'r') as file:
    Corner_pass_length_and_angle_box = file.read()

with open('./figures/Corner_pass_angle_polar.html', 'r') as file:
    Corner_pass_angle_polar = file.read()

with open('./figures/Corner_pass_end_location.html', 'r') as file:
    Corner_pass_end_location = file.read()

with open('./figures/Freeze_frame_game_sitation_goal_vs_nogoal.html', 'r') as file:
    Freeze_frame_game_sitation_goal_vs_nogoal = file.read()

with open('./figures/Freeze_frame_game_sitation_goal_vs_nogoal_polygon_trend.html', 'r') as file:
    Freeze_frame_game_sitation_goal_vs_nogoal_polygon_trend = file.read()

display(HTML(Corner_shot_w_wo_goal_html))

display(HTML(Corner_pass_technique_height))

display(HTML(Corner_pass_w_wo_goal_box))

display(HTML(Corner_pass_length_and_angle_box))

display(HTML(Corner_pass_angle_polar))

display(HTML(Corner_pass_end_location))

display(HTML(Freeze_frame_game_sitation_goal_vs_nogoal))

display(HTML(Freeze_frame_game_sitation_goal_vs_nogoal_polygon_trend))

	id	index	period	timestamp	type	possession	possession_team	play_pattern	...	goalkeeper	dribble	clearance	block	foul_committed	foul_won	50_50	ball_recovery	miscontrol	substitution
0	b074168d-2fb2-4b0a-a9a6-592b37401c7d	1	1	2024-09-11 00:00:00.000	{'id': 35, 'name': 'Starting XI'}	1	{'id': 231, 'name': 'Chievo'}	{'id': 1, 'name': 'Regular Play'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
1	250f4d6f-8cd1-4088-a617-4834082c0a67	2	1	2024-09-11 00:00:00.000	{'id': 35, 'name': 'Starting XI'}	1	{'id': 231, 'name': 'Chievo'}	{'id': 1, 'name': 'Regular Play'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
2	6c8c33b1-3679-4b29-8454-cd79d003dcd0	3	1	2024-09-11 00:00:00.000	{'id': 18, 'name': 'Half Start'}	1	{'id': 231, 'name': 'Chievo'}	{'id': 1, 'name': 'Regular Play'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
3	19810ba9-ec42-43ed-8ac9-7cd441b034e5	4	1	2024-09-11 00:00:00.000	{'id': 18, 'name': 'Half Start'}	1	{'id': 231, 'name': 'Chievo'}	{'id': 1, 'name': 'Regular Play'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4	820f9c95-4382-45a7-bd18-aac9e335f06d	5	1	2024-09-11 00:00:00.327	{'id': 30, 'name': 'Pass'}	2	{'id': 236, 'name': 'Lazio'}	{'id': 9, 'name': 'From Kick Off'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

	id	index	period	timestamp	minute	second	type	possession	possession_team	play_pattern	...	dribble	clearance	block	foul_committed	foul_won	50_50	ball_recovery	miscontrol	substitution	eventcode
174	7010ad8b-3b17-409c-a7b6-df3db59e7815	175	1	2024-09-10 00:03:45.904	3	45	{'id': 30, 'name': 'Pass'}	13	{'id': 236, 'name': 'Lazio'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	corner_pass
537	09b93c63-1d1b-4dd5-9165-5c905da382b1	538	1	2024-09-10 00:13:30.056	13	30	{'id': 30, 'name': 'Pass'}	31	{'id': 236, 'name': 'Lazio'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	corner_pass
771	fb2462f6-4dfd-42eb-8aae-c3caab76c37e	772	1	2024-09-10 00:18:14.391	18	14	{'id': 30, 'name': 'Pass'}	43	{'id': 236, 'name': 'Lazio'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	corner_pass
1198	21dbe7c1-2a84-4016-8a95-fc3d7bd9ddec	1199	1	2024-09-10 00:29:03.123	29	3	{'id': 30, 'name': 'Pass'}	63	{'id': 231, 'name': 'Chievo'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	corner_pass
1654	b9d38208-ea40-46be-9ac8-8fac411574d4	1655	1	2024-09-10 00:39:00.384	39	0	{'id': 30, 'name': 'Pass'}	85	{'id': 236, 'name': 'Lazio'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	corner_pass
2282	733c3345-0366-4613-b5b6-c526dd3503f9	2283	2	2024-09-10 00:09:48.542	54	48	{'id': 30, 'name': 'Pass'}	115	{'id': 236, 'name': 'Lazio'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	corner_pass
2797	f634f512-bafc-410f-b96d-803bda19f2b1	2798	2	2024-09-10 00:21:39.220	66	39	{'id': 30, 'name': 'Pass'}	143	{'id': 231, 'name': 'Chievo'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	corner_pass
3194	6b349d4d-fe86-4a83-ad4b-aec01f349c55	3195	2	2024-09-10 00:32:39.115	77	39	{'id': 30, 'name': 'Pass'}	165	{'id': 236, 'name': 'Lazio'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	corner_pass
3215	791930cd-b628-49d5-b19d-9abfca6e0ae5	3216	2	2024-09-10 00:33:02.513	78	2	{'id': 30, 'name': 'Pass'}	166	{'id': 236, 'name': 'Lazio'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	corner_pass
3442	4b8ac9b4-405b-4235-8e48-bf7be5411e7b	3443	2	2024-09-10 00:39:30.504	84	30	{'id': 30, 'name': 'Pass'}	177	{'id': 236, 'name': 'Lazio'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	corner_pass

	id	goal
1517	938981c0-ec27-432f-ba67-4a5264750294	No
1896	a26bd1d8-8746-4de6-9972-226dbf037123	No
2688	96e67146-48f7-47c4-b956-c1a18bf4e2a1	No
2906	457f342e-94d9-414a-a12b-676b3bdc3793	No
2939	4f811f97-b991-4760-8be4-20e211d91f9a	No
...	...	...
4752	8d13e333-4930-4d4c-a5f7-a998f8c7fc73	No
109	1810f800-e630-4468-9aeb-d8598422e306	Yes
1049	b9f9c48e-9b2d-4758-a47d-1e715eb2f9ee	No
1961	35af67ff-5de3-42e3-aac0-a9a601a4282a	No
2748	3d65bbaf-8449-46bd-83dd-b62bec491ae2	No

	id	index	period	timestamp	minute	second	type	possession	possession_team	play_pattern	...	bad_behaviour	substitution	eventcode	match_id	block	50_50	injury_stoppage	player_off	half_start	half_end
0	491ee6e1-798f-4cf9-ab14-31009ce45cf3	2891	2	2024-09-10 00:39:36.295	84	36	{'id': 16, 'name': 'Shot'}	180	{'id': 216, 'name': 'Getafe'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	goal_via_shot_with_previous_corner	3825894	NaN	NaN	NaN	NaN	NaN	NaN
1	2a4de53e-547b-4485-8972-2aa90a671716	2888	2	2024-09-10 00:39:34.964	84	34	{'id': 30, 'name': 'Pass'}	180	{'id': 216, 'name': 'Getafe'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	corner_pass_for_later_goal	3825894	NaN	NaN	NaN	NaN	NaN	NaN
2	3f0fc8e9-a09f-480a-9396-132e1ca05ec5	256	1	2024-09-10 00:05:46.380	5	46	{'id': 16, 'name': 'Shot'}	16	{'id': 971, 'name': 'Chelsea FCW'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	goal_via_shot_with_previous_corner	7298	NaN	NaN	NaN	NaN	NaN	NaN
3	a8b0ed2d-3c9f-42c4-800e-f944d79e102d	240	1	2024-09-10 00:05:38.140	5	38	{'id': 30, 'name': 'Pass'}	16	{'id': 971, 'name': 'Chelsea FCW'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	corner_pass_for_later_goal	7298	NaN	NaN	NaN	NaN	NaN	NaN
4	10210397-392a-421e-b204-6cb28107b063	2270	2	2024-09-10 00:15:44.419	60	44	{'id': 16, 'name': 'Shot'}	120	{'id': 217, 'name': 'Barcelona'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	goal_via_shot_with_previous_corner	265958	NaN	NaN	NaN	NaN	NaN	NaN

	id	index	period	timestamp	minute	second	type	possession	possession_team	play_pattern	...	foul_won	ball_recovery	bad_behaviour	substitution	block	50_50	injury_stoppage	player_off	half_start	half_end
0	4f711178-73da-4ce6-b073-2c5a23cb9ac7	230	1	2024-09-10 00:07:06.781	7	6	{'id': 30, 'name': 'Pass'}	18	{'id': 1683, 'name': 'Carpi'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
1	921173ba-b094-469a-bb04-8af4140c557b	630	1	2024-09-10 00:20:10.433	20	10	{'id': 30, 'name': 'Pass'}	51	{'id': 226, 'name': 'Hellas Verona'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
2	ba402900-b4bc-47e2-abde-c5750ef839db	841	1	2024-09-10 00:27:37.817	27	37	{'id': 30, 'name': 'Pass'}	61	{'id': 226, 'name': 'Hellas Verona'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
3	a90f151b-a4a6-41f2-815c-a42bfe705ce6	845	1	2024-09-10 00:27:44.774	27	44	{'id': 30, 'name': 'Pass'}	62	{'id': 226, 'name': 'Hellas Verona'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4	3017ff11-eb10-4f1c-8a8e-5bcad01b480f	1504	2	2024-09-10 00:00:58.653	45	58	{'id': 30, 'name': 'Pass'}	117	{'id': 1683, 'name': 'Carpi'}	{'id': 2, 'name': 'From Corner'}	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN

a. Making use of the whole data set of Free Statsbomb Data to help you find general offensive trends.¶

First check for an individual event file and keep only necessary information¶

Prepare multipe functions to extract information about corners linked to a goal event¶

Create new dataset with all necessary information (keep the match-id for each event)¶

Save/Load intermediate dataset consisting of all corner information¶

General overview over event codes¶

Prepare the intermediate dataset for the evaluation¶

Trend 1: Direct shots from the corner (id=61) that hit the goal (id=97)¶

Trend 2: Finding corner passes (id=61) that hit the goal after a shot (id=97) with the corresponding techniques and height of the pass:¶

Trend 3: Finding corner passes (id=61) that hit the goal after a shot (id=97) with their pass length and angle:¶

Trend 4: Finding corner passes (id=61) that hit the goal after a shot (id=97) with their according end location of the ball [x, y]¶

b. Making use of the 360 data for the few matches where they are available, to see if you can find effective defensive set-ups.¶

Preparation of multiple functions for extracting the visible area and the individual positions of the players¶

Visualize for a single freeze frame¶

Load all visible areas and freeze frames and link to the events to find out whether this single freeze frame was under the play pattern of a corner and whether it led to a goal or not¶

Merge this 360-dataset togehter with corner with/without goal events¶

Determine the area in which the players of the attacking team and the defending team are located¶

Save/Load intermediate dataset consisting of all freeze frames information (to csv's)¶

Define a classification model using the features of the 360 dataset (actor, location[x,y]) to determine whether a particular pattern leads to a goal¶

Conclusion on the application of machine learning models¶

c. Summary of the analysis for the Football Set-Piece coach.¶

Offensive recommendation based on Trend 1: Direct shots on goal from the corner¶

Offensive recommendation based on Trend 2: Finding corner passes with a successful shooting technique and appropriate height of the pass¶

Offensive recommendation based on Trend 3: Finding corner passes with successful pass length and angle¶

Offensive recommendation based on Trend 4: Determination of the (x,y) positions from which the corner kick comes to a goal finish¶

Defensive recommendation based on freeze frame analysis 1: Search for the area in which the player is most likely to score goals¶

Defensive recommendation based on freeze frame analysis 2: Search for the running direction in the area where the player is most likely to score goals¶

Defensive recommendation based on freeze frame Machine Learning-based analysis: Machine learning for optimal positioning of players¶

Concluding remarks:¶

	id	Corner-pass technique	Corner-pass height	Pass length	Pass angle	goal
1	2a4de53e-547b-4485-8972-2aa90a671716	Inswinging	High Pass	32.339450	1.663696	Yes
5	1b1cf446-bd02-4f88-bd93-d0a2a7efadfb	Outswinging	High Pass	44.537174	-1.797259	Yes
7	53020c5a-28da-4292-b9c5-e0064303a5e3	Straight	High Pass	37.258020	1.831399	Yes
9	f9c435a5-21a5-49b2-86a9-47bb0db3e1a3	Inswinging	High Pass	36.122154	1.681759	Yes
11	55340a86-9415-4a28-80d2-a5bfe99a31d6	Inswinging	High Pass	50.243904	1.646500	Yes

	event_uuid	visible_area	freeze_frame
0	5a1bb2e7-fde1-4e3d-a331-2f6d4698e36e	[12.307859690377, 80.0, 41.9817181390851, 0.0,...	[{'teammate': True, 'actor': False, 'keeper': ...
1	8030d926-c5fb-4abf-af7c-216eb4df13d9	[11.862200102872, 80.0, 41.7940694041057, 0.0,...	[{'teammate': True, 'actor': False, 'keeper': ...
2	14f4c9f4-33cd-4ca1-97a7-8d4d0d209609	[11.862200102872, 80.0, 41.7940694041057, 0.0,...	[{'teammate': True, 'actor': False, 'keeper': ...
3	c92c6e1c-c6a7-4018-8321-093241aa6596	[23.6166850551218, 80.0, 47.4167371969312, 0.0...	[{'teammate': True, 'actor': False, 'keeper': ...
4	594b9080-ff52-4a3c-ad4b-69da13403cbd	[34.7462953710152, 80.0, 49.9796978695372, 7.0...	[{'teammate': True, 'actor': False, 'keeper': ...

	event_uuid	visible_area	freeze_frame
42	2439c812-71f9-4fc1-864b-61fe7892add7	[0.0, 80.0, 0.0, 61.6619136535695, 30.32469024...	[{'teammate': True, 'actor': True, 'keeper': T...
43	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	[22.4985076696792, 80.0, 42.3872165649152, 14....	[{'teammate': True, 'actor': False, 'keeper': ...

	event_uuid	visible_area_x	visible_area_y
0	2439c812-71f9-4fc1-864b-61fe7892add7	0.0	80.0
1	2439c812-71f9-4fc1-864b-61fe7892add7	0.0	61.661914
2	2439c812-71f9-4fc1-864b-61fe7892add7	30.32469	9.006613
3	2439c812-71f9-4fc1-864b-61fe7892add7	61.342665	18.390359
4	2439c812-71f9-4fc1-864b-61fe7892add7	61.077709	80.0
5	2439c812-71f9-4fc1-864b-61fe7892add7	0.0	80.0
6	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	22.498508	80.0
7	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	42.387217	14.334681
8	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	73.192729	14.556697
9	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	90.656039	80.0
10	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	22.498508	80.0

	event_uuid	teammate	actor	keeper	location_x	location_y	teammate_category	actor_category	keeper_category
0	2439c812-71f9-4fc1-864b-61fe7892add7	True	True	True	23.5	52.0	teammate	actor	keeper
1	2439c812-71f9-4fc1-864b-61fe7892add7	True	False	False	27.298834	65.009051	teammate	non-actor	player
2	2439c812-71f9-4fc1-864b-61fe7892add7	True	False	False	30.240036	42.367776	teammate	non-actor	player
3	2439c812-71f9-4fc1-864b-61fe7892add7	True	False	False	34.993533	19.526347	teammate	non-actor	player
4	2439c812-71f9-4fc1-864b-61fe7892add7	False	False	False	36.08542	44.067301	opponent	non-actor	player
5	2439c812-71f9-4fc1-864b-61fe7892add7	False	False	False	39.273706	61.974411	opponent	non-actor	player
6	2439c812-71f9-4fc1-864b-61fe7892add7	True	False	False	46.734633	42.927346	teammate	non-actor	player
7	2439c812-71f9-4fc1-864b-61fe7892add7	False	False	False	52.000885	33.647706	opponent	non-actor	player
8	2439c812-71f9-4fc1-864b-61fe7892add7	False	False	False	53.870975	21.787159	opponent	non-actor	player
9	2439c812-71f9-4fc1-864b-61fe7892add7	True	False	False	57.090807	83.358643	teammate	non-actor	player
10	2439c812-71f9-4fc1-864b-61fe7892add7	True	False	False	58.514414	28.945314	teammate	non-actor	player
11	2439c812-71f9-4fc1-864b-61fe7892add7	False	False	False	58.737428	58.680872	opponent	non-actor	player
12	2439c812-71f9-4fc1-864b-61fe7892add7	False	False	False	60.865629	40.82181	opponent	non-actor	player
13	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	True	False	False	27.376394	69.065884	teammate	non-actor	player
14	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	False	False	False	38.614621	64.050564	opponent	non-actor	player
15	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	True	False	False	44.297048	49.999829	teammate	non-actor	player
16	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	False	False	False	50.522513	33.374152	opponent	non-actor	player
17	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	True	True	False	50.799999	78.0	teammate	actor	player
18	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	False	False	False	52.014459	20.242657	opponent	non-actor	player
19	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	True	False	False	53.855895	31.897164	teammate	non-actor	player
20	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	False	False	False	54.155644	69.045004	opponent	non-actor	player
21	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	True	False	False	59.859808	60.291	teammate	non-actor	player
22	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	False	False	False	60.530639	39.189189	opponent	non-actor	player
23	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	False	False	False	63.505657	63.382132	opponent	non-actor	player
24	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	True	False	False	64.853683	32.322059	teammate	non-actor	player
25	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	False	False	False	66.626958	31.054405	opponent	non-actor	player
26	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	False	False	False	70.858948	43.256373	opponent	non-actor	player
27	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	True	False	False	70.95654	52.353921	teammate	non-actor	player
28	aebf05f3-75d3-4191-9a8c-08f07d9eb4b5	False	False	False	71.867798	51.760465	opponent	non-actor	player

	event_uuid	visible_area_x	visible_area_y
0	75d6cc25-b03b-44e0-9c50-99a7e3c47315	29.574168	80.0
1	75d6cc25-b03b-44e0-9c50-99a7e3c47315	47.799207	0.0
2	75d6cc25-b03b-44e0-9c50-99a7e3c47315	74.417607	0.0
3	75d6cc25-b03b-44e0-9c50-99a7e3c47315	94.482156	80.0
4	75d6cc25-b03b-44e0-9c50-99a7e3c47315	29.574168	80.0
...	...	...	...
25603	29761d25-06f1-45c1-b72e-a01c14633e84	15.981348	0.0
25604	29761d25-06f1-45c1-b72e-a01c14633e84	17.39688	0.0
25605	29761d25-06f1-45c1-b72e-a01c14633e84	41.230627	16.308263
25606	29761d25-06f1-45c1-b72e-a01c14633e84	17.902746	80.0
25607	29761d25-06f1-45c1-b72e-a01c14633e84	0.0	80.0

	event_uuid	teammate	actor	keeper	location_x	location_y	teammate_category	actor_category	keeper_category	goal	marker_size
6815	ca18ba8a-13ce-4287-aa81-0b02a0e2e4df	False	False	False	95.251147	43.113602	opponent	non-actor	player	No	0.07
6816	ca18ba8a-13ce-4287-aa81-0b02a0e2e4df	False	False	False	98.217912	28.00216	opponent	non-actor	player	No	0.07
6817	ca18ba8a-13ce-4287-aa81-0b02a0e2e4df	True	False	False	101.821481	38.905794	teammate	non-actor	player	No	0.07
6818	ca18ba8a-13ce-4287-aa81-0b02a0e2e4df	False	False	False	106.777983	36.508244	opponent	non-actor	player	No	0.07
6819	ca18ba8a-13ce-4287-aa81-0b02a0e2e4df	False	False	False	107.244791	41.116926	opponent	non-actor	player	No	0.07
...	...	...	...	...	...	...	...	...	...	...	...
14402525	8d13e333-4930-4d4c-a5f7-a998f8c7fc73	False	False	False	116.283493	40.831504	opponent	non-actor	player	No	0.07
14402526	8d13e333-4930-4d4c-a5f7-a998f8c7fc73	False	False	False	116.465836	31.71431	opponent	non-actor	player	No	0.07
14402527	8d13e333-4930-4d4c-a5f7-a998f8c7fc73	False	False	False	117.268123	34.04374	opponent	non-actor	player	No	0.07
14402528	8d13e333-4930-4d4c-a5f7-a998f8c7fc73	False	False	True	118.086067	39.303187	opponent	non-actor	keeper	No	0.07
14402529	8d13e333-4930-4d4c-a5f7-a998f8c7fc73	True	False	False	118.15584	0.822093	teammate	non-actor	player	No	0.07